Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
VALUATION OF MORTGAGE-BACKED SECURITIES
Rui Huang Bachelor of Finance, Shanghai Institute of Foreign Trade, 2004
PROJECT SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS
In the Faculty
of Business Administration
Financial Risk Management
O Rui Huang 2006
SIMON FRASER UNIVERSITY
Summer 2006
All rights reserved. This work may not be reproduced in whole or in part, by photocopy or other means, without permission of the author.
APPROVAL
Name:
Degree:
Title of Project:
Rui Huang
Master of Arts
Valuation of Mortgage-backed Securities
Supervisory Committee:
Geoffrey Poitras Senior Supervisor Professor of Business Administration
Date Approved:
Chris Veld Supervisor Associate Professor of Business Administration
SIMON FRASER UNIVERSITY~I brary
DECLARATION OF PARTIAL COPYRIGHT LICENCE
The author, whose copyright is declared on the title page of this work, has granted to Simon Fraser University the right to lend this thesis, project or extended essay to users of the Simon Fraser University Library, and to make partial or single copies only for such users or in response to a request from the library of any other university, or other educational institution, on its own behalf or for one of its users.
The author has further granted permission to Simon Fraser University to keep or make a digital copy for use in its circulating collection, and, without changing the content, to translate the thesislproject or extended essays, if technically possible, to any medium or format for the purpose of preservation of the digital work.
The author has further agreed that permission for multiple copying of this work for scholarly purposes may be granted by either the author or the Dean of Graduate Studies.
It is understood that copying or publication of this work for financial gain shall not be allowed without the author's written permission.
Permission for public performance, or limited permission for private scholarly use, of any multimedia materials forming part of this work, may have been granted by the author. This information may be found on the separately catalogued multimedia material and in the signed Partial Copyright Licence.
The original Partial Copyright Licence attesting to these terms, and signed by this author, may be found in the original bound copy of this work, retained in the Simon Fraser University Archive.
Simon Fraser University Library Burnaby, BC, Canada
ABSTRACT
This paper attempts to provide a method for the valuation of mortgage-backed securities
(MBS).
Specifically, Modified Goldman Sachs model is selected to describe mortgagors'
prepayment behaviour, which takes account of mortgage's refinancing incentive, aging effect,
month effect and burnout effect. In this study, I will use above model to price a pass-through
MBS and a plain vanilla MBS respectively.
Furthermore, I use scenarios testing to discuss how MBS's price and standard deviation
(risk) change if mortgage property and model parameters change.
Keywords: Mortgage-backed Securities; Prepayment Model; Morte Carlo Simulation
ACKNOWLEDGEMENTS
I would like to thank Professor Geoffiey Poitras for his encouragement and continued
support for my paper.
I also wish to thank Professor Robbie Jones for providing his valuable treasure, sharing
his knowledge with me.
TABLE OF CONTENTS
. . Approval .......................................................................................................................................... 11
... Abstract .................................................................................................................................. 111
Acknowledgements ........................................................................................................................ iv
........................................................................................................................... Table of Contents v ..
List of Figures ............................................................................................................................... VII
.. ................................................................................................................................ List of Tables VII
.............................................................................................................................. Introduction 1
...................................................................................... Risk of Mortgage-backed Securities 2 ..................................................................................................................... 2.1 Refinancing 2
...................................................................................... 2.2 House Turnover or House Sales 2 .................................................................................................. 2.3 Involuntary Prepayment 2
Mortgage Modelling Overview ............................................................................................... 4 .............................................................................................................. 3.1 Structural Model 4
..................................................................................................... 3.2 Reduced Form Model 5
Prepayment Model ................................................................................................................... 7 ...................................................................................................... 4.1 12-year Life Method 7
4.2 FHA Experience Method ................................................................................................ 7 ................................................................. 4.3 CPR (Conditional Prepayment Rate) Method 7
............................................... 4.3.1 PSA (Public Security Association) Standard Method 8 ..................................... 4.3.2 Schwartz and Torous 's Proportional Hazard Model (PHM) 8
................................................ 4.3.3 Richard and Roll's Modified Goldman Sachs Model 9
............................................................................................................... Interest Rate Model 12
............................................................................................................... 5.1 Vasicek Model 12 ..................................................................................................................... 5.2 CIR Model 12
Plain Vanilla MBS ............................................................................................................. 14
MBS Cash Flows .................................................................................................................... 15
........................................................................................................................... MBS Pricing 19 ................................................................................................ 8.1 Pricing a Mortgage Pool 19 ............................................................................................... 8.1.1 Change Coupon Rate 25
........................................................................................... 8.1.2 Change Mortgage Term 26
.......................................................................................... 8.1.3 Change CIR Parameters -26
......................................................................................... 8.2 Pricing a Plain Vanilla MBS 28 8.2.1 Change Mortgage Term ........................................................................................... 29
.......................................................................................... 8.2.2 Change CIR Parameters 31
.............................................................................................................................. 9 Conclusion 33
Reference List ............................................................................................................................... 34
LIST OF FIGURES
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Month Factor .............................................................................................................. 11
Cash Flow Pattern for Regular Bonds ................................................................... 15
Cash Flow Pattern for MBS ..................................................................................... 16
Computation Process .............................................................................................. 19
One Interest Rate Path ................................................................................................ 21
One R Path .................................................................................................................. 23
One CPR Path ............................................................................................................ 2 4
............................................. Positive Relationship Between Price and Coupon Rate 26
LIST OF TABLES
Table 1
Table 2
Table 3
Table 4
Table 5
Table 6
Table 7
Table 8
Table 9
Mortgage Pool ............................................................................................................ 20
Estimated Results for Mortgage Pool (Different Coupon Rate) ................................. 25
Estimated Results for Mortgage Pool (Different Mortgage Term) ............................. 26
Estimated Results for Mortgage Pool (Different Mean Reverse Speed) .................... 27
Estimated Results for Mortgage Pool (Different Interest Volatility) .......................... 27
CMO Structure ........................................................................................................... 28
Tranche Price (Different Mortgage Term) .................................................................. 29
Tranche Price (Different Mean Reverse Speed) ........................................................ 3 1
Tranche Price (Different Interest Volatility) .............................................................. 32
vii
INTRODUCTION
Mortgage-backed securities (MBS) are investment instruments that represent ownership
of a group of mortgages. Usually, financial institutions treat MBS as a method to fund the capital
pool brought by lending mortgages.
Up to 2006, MBS market is one of the fastest growing, as well as one of the largest
financial markets in the United States. There are now three major agencies in U.S. issuing MBS -
Ginnie Mae (the Government National Mortgage Association, or GNMA), Fannie Mae (the
Federal National Mortgage Association, or FNMA), and Freddie Mac (the Federal Home Loan
Mortgage Corporation, or FHLMC). These agencies support the secondary market by issuing an
MBS in exchange for pools of mortgages from lenders. This ensures that lenders have funds to
make additional loans. In general, MBS issued through these three government-supported
agencies are considered to have no credit or negligible credit risk.
The paper will be structured in this way. In section 2, we will look at risks behind
mortgage-backed securities and factors causing those risks. Mortgage pricing models developed
in the recent past will be presented in section 3. In section 4, we will discuss some prepayment
models and Modified Goldman Sachs model is chosen to price MBS. In section 5, we will choose
the interest rate process. Then in section 6, we will introduce the specific CMO structure we will
price later - plain vanilla. In the next section, section 7, MBS cash flow streams are discussed. In
section 8, the computation process is detailed, and we will present the numerical results. Section 9
will be the last section where the conclusion is presented.
2 RISK OF MORTGAGE-BACKED SECURITIES
The biggest uncertainty of mortgage-backed securities is the prepayment behaviour. In
U.S., mortgagors could prepay their loans at any time with no or only little penalty. The above
earlier payments lead to unstable fiture cash flow, thus uncertainty of MBS fair price today. To
some extent, mortgages can be treated as a callable bond and people could exercise their option
any time they like. So before we begin to price MBS, What we want to know is at what time
those borrowers will exercise the call option and what the vaiue of that embedded option is.
Roughly, there are three reasons causing the prepayment behaviours.
2.1 Refinancing
When interest rates decline, mortgagors originally signed the higher rate contract would
like to refinance through entering a new mortgage contract, thus to take advantage of the
prevailing low interest rates.
2.2 House Turnover or House Sales
Many mortgage contracts specify the settlement of remaining loan before selling out the
house. So people would like to prepay before mortgage mahrity date if they want to sell the
house.
2.3 Involuntary Prepayment
Prepayment could occur unexpectedly. Such early termination events include marriage,
divorce, death and other extraordinary events.
Lots of studies then report that several factors affect those behaviours - interest rate as
mentioned above, mortgage seasoning, borrowers' heteroscedasticity, overall economy, burnout
effect, even geographic diversity and schooling time.. . . It seems to be a tough work if we want to
construct a model with these multitudinous factors, but still, several models have been built
through chasing one or several above variables.
3 MORTGAGE MODELLING OVERVIEW
Many literatures on mortgage valuation have been published in the past 20 years.
3.1 Structural Model
Kenneth Dunn was the first to introduce an MBS model and then a series of the papers
with McConnell in 198 1. They treat prepayment as a call option and their model assumes optimal
refinancing decisions. Although these models consistently link valuation and prepayment, their
prepayment predictions assume an instant prepayment action by all the mortgagors when interest
rates hit a threshold. Tirnmis, Dunn and Spatt, and Johnston and Van Drunen improved this
model by incorporating transaction cost or other frictions to explain refinancing decisions.
Noticing that some people may not exercise their options when interest rate is low, in the
following years, this option-based method has been improved by adding so called option
probability model or other determined variables. Downing, Stanton and Wallance established a 2
factor model use both interest rate and house price as a joint threshold and option probability
model by a hazard rate fimction. Other researchers, like Kariya, Ushiyama, Pliska construct a 3
factor model using treasury yield, mortgage rate and house price.
The general idea of the structural model begins with the "public believed" movement of
determined variables, usually interest rate and house price. Then Ito's lemma is used to get a
"pure" PDE equation and prepayment noises are added to compute the "real" PDE equation.
Finally we solve this PDE equation with boundary condition.
For example, in 2 factor model, where we assume house price and interest move as
follows.
dH = (u, - s)Hdt + o, H ~ Z , ~
Where s represents risk adjusted drift when we applied to the risk neutral world.
dr = y(B - r)dt + or &dzr CIR (1985)
We could get PDE for mortgage X(r,H,t) (That is the core idea of the structural model)
Combining this pure equation with boundary condition and option exercise probability
model, we could get the final numerical answer. Today, different finite backward methods are
used to solve this problem.
It seems to be a compromise of reality, since option-base technique is a "perfect,
classical, non-arbitrage" thought that investors act optimally, while option exercise probability
model is generally a result of irrational exercise behaviours.
3.2 Reduced Form Model
Since above theoretical process asks for very complex calculation, practitioners in
industry refer to find other comparatively easy-to-understand solution. Reduced form model
assumes option exercise is built in a prepayment model and evaluates mortgage price by
discounting future cash flow to its present value. It is easy to modify and needs less computation.
No doubt when those academic geniuses were hired by those big investing companies or financial
institutions they all choose to use this method.
The basic idea of this method is that the price of MBS today should be the present value
of total cash inflow which could be earned in the future.
Here we assume a 30-year mortgage, v l , v2,. . . are explainable variables, 21, 22,. . . are
implied yield.
Usually, cash flows are determined by a prepayment model and Monte Carlo simulation
is used to simulate interest process and cash flows. In the paper we prefer this method and will
use it to compute MBS pool, but first, let us look at some different prepayment models.
4 PREPAYMENT MODEL
4.1 12-year Life Method
According to the material from Federal Housing Administration (FHA), a 30-year
mortgage loan will be hlly paid in the 12th year, so GNMA obtain MBS price by assuming that
the 30-year mortgage will have normal payment in the first 143rd months then repay the all the
balance remaining in the 144th month. This coarse method is quickly substituted by other more
complex method since there is evidence that the prepayment model is determined by a number of
different factors.
4.2 FHA Experience Method
FHA built a yearly survivorship table according to its historical 30-year mortgage data.
Prepayment rate is lower during initial loan period, then increases, and reaches its crest between
the 5th and 8th year, afterwards the number drops. The advantage of this method is prepayment
could vary every year. The disadvantage is that those numbers only reflect the prepayment rate in
a specified period and do not take interest rate into account.
4.3 CPR (Conditional Prepayment Rate) Method
There has been a proliferation of CPR models in financial literatures. These models
include those by PSA (Public Security Association), Asay, Guillaurne and Mattu (1987), Chinloy
(1 989), Davidson, Herskovitz and Van Drunen (1 989), Giliberto and Thibodeau (1 989), Richard
and Roll (1989), Schwartz and Torous (1989). Among those models, PSA method is an "easy to
implement" one and a learning objective in recent CFA curriculum. The last two models are more
popular and referenced time after time in the recent papers, so we will discuss them in detailed.
4.3.1 PSA (Public Security Association) Standard Method
Public Security Association supposes the prepayment rate in the first year is 0.2% and
this rate grows by 0.2% every month until the 3oth month. It could be formulated by:
t CPR =6%*- when t<30
30
CPR = 6% when t>=30
PSA treats above formula as its 100% mortgage prepayment benchmark. For different
PSA mortgage, for example, 150% PSA, you just need to multiply 1.5 on the above formula.
4.3.2 Schwartz and Torous's Proportional Hazard Model (PHM)
They use the following proportional hazard model to compute CPR.
CPR = ~ ( t ; v, ,8) = r,, (t; r, p) * exp(pv)
v is the vector of determined factors; P is the vector of coefficients.
Particularly, they use the following four factors.
Here v, represents the difference between mortgage rate and current long term treasury
rate. c is mortgage rate; 1 is long term treasury rate at time t-s, where s is the lag factor. s=3 is
used in this model.
Here v2 is just cubed equation of vl, representing the accelerate effect when treasury rate
is sufficiently lower than the mortgage rate.
v, (t) = ln(BAL(t) l BAL * (t))
v3 represents the burnout effect where BAL(t) is the mortgage pool outstanding at time t
and BAL*(~) is pool outstanding in the absence of prepayment.
1 when t = May through August
0 when t = September through April
v4 represents the seasoning factor on prepayment decision.
zo (t; r, p) is baseline function which explain the CPR when factor vector v=O, and they
use log-logistic function to describe .no
This model is intuitive because it divides the mortgage prepayment risk into two parts.
The first part, 7~ (also called aging effect) tells us that mortgage will suffer a baseline risk at any
time of mortgage period. The second part, exp (pv) represents all the mortgage specified risk.
4.3.3 Richard and Roll's Modified Goldman Sachs Model
The model is first constructed when Richard and Roll were hired by Goldman Sachs as
financial advisors and then developed by OTS (Office of thrift supervision). They point out that
CPR is determined by the following four factors and measure CPR as the product of those four
factors:
CPR = (Refinancing Incentive)*(Seasoning Factor)*(Month Factor)*(Pool Burnout Factor)
Where CPR is the conditional annual prepayment rate.
Refinancing incentive in their model could be measured by the ratio or the spread of
weighted average mortgage coupon rate and mortgage refinancing rate.
RI = a + b * arctan(c + d(WAC - R))
Seasoning reflects the observation that newer loans tend to prepay slower than older or
"seasoned" loans. This factor follows the rationale behind the PSA standard prepayment model.
This industry convention adopted by the Public Securities Association models mortgage
prepayment rates as increasing linearly from 0.2% CPR at issue to 6% CPR at thirty months and
then remaining constant. In the paper, the formula for Seasoning is:
Month factor measures different prepayment behaviours during 1 year. It is believed that
prepayments peak in the summer and decrease in the winter. One of the sources of prepayments
due to seasonality is housing turnover. This could be due to weather and school schedules. In this
paper, the monthly parameters for ith month were taken from the figure in Richard and Roll's
paper (1989).
Figure 1 Month Factor
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Source: data from Richard and Roll, 1989
Burnout takes into account the tendency for prepayment to diminish over time, even
when refinancing incentives are favourable. Richard and Roll try to quantify burnout by
measuring how much the option to prepay has been deep in-the-money since the pool was issued.
They suggest the more the prepayment option has been deep in-the-money, the more burned out
the pool is, and the smaller prepayments are, all other things being equal. The Burnout factor is
calculated as follows:
Where B(t) is the mortgage balance in the beginning of month t, B(0) is the initial
mortgage balance.
5 INTEREST RATE MODEL
Interest rate is notoriously difficult to predict. Recent interest model could be categorized
into two groups - equilibrium models and non-arbitrage model. The main difference of the two
lies in the different description of interest rate drift. In equilibrium model, drift is not set as a
fknction of time while non-arbitrage model takes it into conqideration. In the paper we will
introduce non-arbitrage models which are more commonly used in MBS evaluation.
5.1 Vasicek Model
Vasicek (1 977) proposes a kind of interest rate model
dr = a(b - r)dt + a l z (Vasicek Model)
Where a is instantaneous drift speed, b is long term equilibrium interest rate, o is the
instantaneous standard deviation and dz is a Wiener process
The Advantage of this model is the built-in mean-reversion idea in the equation. When
short-term interest rate deviates from long time equilibrium level b, it will then be reversed to b at
the speed of a.
5.2 CIR Model
Cox, Ingersoll and Ross (1985) improve Vasicek's model by modifying instantaneous
standard deviation from o to 06
(CIR Model)
This model follows the idea of mean-reversion phenomenon of interest rate. Besides, it
will not generate negative interest rate. So I pick this model to build our interest rate path.
Parameters in our interest rate model are described in the "M~rtgage Pool Pricing" section.
6 PLAIN VANILLA MBS
Mortgage pool is not expected to be sold in one pile. Usually, those agencies will issue
securities by redirecting the cash flows of mortgages in order to attract different investors. They
redirect cash flows in such way that one classes of bond will be only exposed to one form of
prepayment risk. When the cash flows of mortgage-backed products are redistributed to different
bond classes, the resulting securities are called collateralized mortgage obligations (CMOs). Plain
Vanilla CMOs, also called Sequential Pay CMOs, is the most basic class within CMOS structure.
This class reallocates collateral principal payments sequentially to a series of bonds, namely
A,B,C,D.. . . Bond A will receive principal payments, including prepayments first. When it is fully
retired; then principal payments are redirected to the next sequential class, bond B. This process
continues until the last sequential class is fully retired. While one class is receiving principal
payments, the other existing classes only receive monthly interest payments at their coupon rate
on their principal.
7 MBS CASH FLOWS
As mentioned above, mortgage is just like an ordinary coupon bond (without final
principal payment) if there is no prepayment. The interest part and principal part in every
schedule payment under no prepayment assumption will follow the pattern in figure 2.
Figure 2 Cash Flow Pattern for Regular Bonds
I I
0 10 2 0 30 years
Source: www.riskglossaly.com
But when prepayments happen, the interest part and the principal part will be uncertain. It
may follow the pattern below.
Figure 3 Cash Flow Pattern for MBS
servicing fee
principal
1 interest
years
Source: www.riskglossay.com
In Fabozzi's paper, the cash flows can be calculated using the following formula:
Where MP(t) is the scheduled mortgage payment for period t, B(t) is the mortgage
balance remaining at t, WAC is weighted average coupon rate and WAM is the weighted average
maturity of mortgage.
WAC IP(t) = B(t) * -
12
Where P(t) is the interest payment at t.
Where SIJ(t) is the scheduled principal payment for period t;
PP(t) = SMM ( t ) * (B( t ) - SP(t))
Where PP(t) is the principal prepayment for period t and SMM is single month mortality
and computed as
SMM ( t ) = 1 - d m
We could calculate cash flow using CPR(t), the conditional prepayment rate, which was
derived from our prepayment model. Once CPR(t) is known, everything else can be calculated
and total cash flow at time t CF(t) will equal:
Then we could compute the price of mortgage pool hy discounting monthly cash flow.
P, = CF(1) + CF(2) + ... + CF (360) (1 + S, + OAS) ( 1 + S, + O A S ) ~ (1 + s,,, + 0 ~ s ) ~ ~ ~
Where S is the spot yield curve, computed by our simulated interest rate process.
S,, = d ( l + r l ) ( l + r 2 ) ...( l + r n ) -1
OAS is the implied spread from historic MBS data that makes the computed price of
mortgage pool equal to its market price. In this paper OAS is set to zero, since we only want to
compare pure outcomes between different classes.
The only difference between an entire mortgage pool and a Plain Vanilla MBS is that all
the principal payments and prepayments will be used to retire first sequential bonds, say Bond A.
The principal of Bond B will begin to decline until Bond A is fully amortized, so the calculation
will be as follows if we assume there are two classes, Bond A and Bond B.
W I ) + CF(2) + ... + CF ( t 1) PBondA = ( l + s , +OAS) (1+s2 +OAS)* ( 1 + s,, + 0 ~ s ) ' ~
Where t l is the time Bond A is fully retired and cash flow CF(t) will be
Where IPA(t) is the interest payment according to the remaining balance of bond A at
time t, SPT(t) and PPT(t) is the schedule principal payment and prepayment computed from the
total mortgage pool.
IP + IP + ... + CF (t 1) PBondB = + ... + CF(4
(1 + S, + OAS) (1 + S, + OAS)' (1 + sf, + OAS)" ( 1 + st + OAS)'
where IP is fixed interest payment of bond B when bond A is in the process of retiring.
CF(t1) = SPT(t1) + PPT(t1) + IPB(t1) is the first cash flow of bond B that contains
prepayment and schedule principal payment.
8 MBS PRICING
After every computation processes are introduced, we could then price MBS through the
following flow chart.
Figure 4 Computation Process
Interest Process L-h Prepayment Model c Cash Flow Stream -
MBS Price I 8.1 Pricing a Mortgage Pool
In this paper, I assume a mortgage-backed security with collateral value equal to
1000000. Its principal values, coupon rates, issue date, maturity are described as follows
(assuming Gross coupon rate = Net coupon rate, Settle date = Issue date).
Table 1 Mortgage Pool
Principal
First, I use CIR model to generate interest rate process. Particularly, I use the
discretization form of CIR model, an AR(1) model to generate interest rate path.
$1,000,000
Above AR(1) model is said to have the same mean and stationary variance if
Issue Date
2 2 w2 ( = e-" and C T ~ = CT - 2a
Jan 01, 2000
The long-run average, b, was set to 7.75%. a was set to 0.1 and the volatility term, o, was
set to 0.0225. These parameters are picked from Buser and Hendershott's article (1984). The
starting interest rate was set to 10%. Figure 5 plots one path of the interest rate process used in
my simulation.
Mortgage Term Coupon Rate
30 year 8%
Figure 5 One Interest Rate Path
Here I choose Modified Goldman Sachs prepayment model to valuate MBS. Particularly,
refinancing incentive is computed using following formula and parameters are picked from
Akesson and Lehoczky's simulation paper (2000).
RT = a + b * arctan(c + d(WAC - R))
WAC is weight average coupon rate
R(t,T) is calculated using following method.
According to Hull (2000), if interest rate follows CIR movement,
dr = a(b - r)dt + o&dz
then bond price P(t,T) will equal to
P ( ~ , T ) = e-R(tJ ')*(T-t) = A ( ~ , T ) ~ - B ( ~ . T ) * ~
2(e7'(T-t) - 1) where B(t, T) =
( y + a)(eY(T-" - 1 ) + 2 y
with y = J a 2 + 20'
Figure 6 plots one path of the R process used in my simulation.
Figure 6 One R Path
0.085
Recall that RI = a + b * arctan(c + d(WAC - R)) and CPR = (Refinancing
Incentive)*(Seasoning Factor)*(Month Factor)*(Pool Burnout Factor)
Figure 7 then plots one path of the CPR process used in my simulation.
Figure 7 One CPR Path
I ran 1000 simulations to get the MBS price. Making it simple, I set the market price of
the collateral and the bonds to their respective present values, assuming OAS to equal 0 and
observe what will happen if we change the parameters of the model.
8.1.1 Change Coupon Rate
Table 2 Estimated Results for Mortgage Pool (Different Coupon Rate)
Mortgage Pool 1
Table 2 above presents the price and standard deviation of the mortgage pool. Here I set
initial interest rate to 5%, 8% and lo%, one higher than, one lower than and one close to the long-
term rate and pick different coupon rate (4%, 6%, 8%, 10% and 12%). As coupon rate increases,
prepayment becomes more probable, thus the price of mortgage pool increase. Compared to
initial interest rate lo%, pool prices at 5% interest rate level zxe generally higher, reflecting a
more frequent prepayment behaviour in the beginning period when interest rates have not yet
converged to the long term rate, 7.75%. Figure 8 below presents the relationship between pool
price and coupon rate at two initial interest level (5% and 10%).
Coupon rate
Pool Std
4%
804,809
784,392
-
Pool Price r(0)=5%
r(0)=8%
r(0)=5%
r(0)=8%
6%
927,047
904,781
26,203
25,597
8%
1,054,560
1 ,O3U,i'2l
27,241
25,969
10%
1 ,I 88,868
1 , I 64,248
12%
1,325,394
1,298,989
27,707
28,395
27,684
27,883
27,633
28,724
Figure 8 Positive Relationship Between Price and Coupon Rate
I
Coupon ra te
8.1.2 Change Mortgage Term
Table 3 Estimated Results for Mortgage Pool (Different Mortgage Term)
Table 3 above presents the price and standard deviation of the mortgage pool in different
Mortgage Pool with coupon rate 8%
mortgage term. We could see from the table that both price and standard deviation accrue when
we increase the mortgage term. It can be indicated that uncertainty of prepayment increases when
Mortgage Term
the mortgage term is much longer.
10
1,038,228
1,016,757
1,002,264
17,701
Pool Price
Pool Std
8.1.3 Change CIR Parameters
r(0)=5%
r(0)=8%
r(0)=10%
r(0)=5%
Interest rate played an important role when we valuate the mortgage-backed securities.
15
1,044,449
1,022,209
1,008,403
21,466
Here I present the MBS results corresponding to different mean reverse speed (a=0.04, a=0.07,
a=0.1, a=0.13, a=0.16) and different interest rate volatility (a=0.0025, ~ ~ 0 . 0 1 2 5 , ~ 0 . 0 2 2 5 ,
20
1,048,359
1,026,044
1,011,768
22,594
0=0.0325). The results are shown in table 4 and table 5 respectively.
25
1,050,507
1,029,940
1,013,658
24,728
30
1,054,560
1,032,925
1,016,203
25,885
Table 4 Estimated Results for Mortgage Pool (Different Mean Reverse Speed)
Mortgage Pool with coupon rate 8%
Mean Reverse Speed 1 0.04 1 0.07 1 0.10 1 0.13 1 0.16
Pool Price
Table
Pool Std
Estimated Results for Mortgage Pool (Different Interest Volatility)
Mortgage Pool with coupon rate 8%
r(0)=5%
r(0)=10%
I Interest Volatility 1 0.0025 1 0.0125 1 0.0225 r0.0325 1
r(0)=5%
r(0)=10%
I Pool Price I r(0)=5% 1 1,056,620 1 1,055,040 1 1,054,560 1 1,051,589 1
1,059,097
988,296
I Pool Std ( r(0)=5% 1 3,107 1 15.625 1 27,707 ( 35,949 1
41,170
50,562
It seems CIR parameters have a huge effect on the pool standard deviation. Pool standard
deviation soars if we decrease mean reverse speed and increase interest volatility. What is worth
mentioning is that, when initial rate is lo%, pool prices increase with the increase of mean
reverse speed; It changes in opposite direction when initial rate is set to 5%. It could be explained
by the intuition that, at a lower initial rate, more prepayment behaviours are expected at lower
mean reverse speed than at higher reverse speed, it will increase the cash flows in the first several
years, thus increase the pool price. On the other hand, at a higher rate, more prepayment
behaviours are expected under higher mean reverse speed, so the pool price will be higher.
1,056,553
1,007,704
31,736
34,147
1,054,560
1,016,203
27,707
27,903
1,050,852
1,020,210
1,049,826
1,023,411
21,885
22,183
18,602
18,239
8.2 Pricing a Plain Vanilla MBS
After we analyze the mortgage pool, let us look at a simple CMO structure. Table 6
below presents a four tranche Plain Vanilla MBS. Coupon rates in every tranche are picked to
make the sum of tranche coupon payments equal to pool coupon payments.
Table 6 CMO Structure
Principal Issue Date Mortgage 1 Term
1000000 Jan 01, 30 year 2000
- Tranche 350000 A
Tranche 150000 D
I Total Coupon Payment
Coupon Rate
Coupon Payment
80000
The main difference is we now have interest payment for every tranche and monthly
principle payments are paid in sequence. Let IP(t)i denote the interest payments for tranche i in
month t.
WA Ci IP(t) = Bi ( t ) * -
12
I ran 500 simulations to get tranche prices and I will repeat the similar process (set initial
rate to 5% and 10%) to analyze the prices and risks under different situations.
8.2.1 Change Mortgage Term
Table 7 Tranche Price (Different Mortgage Term)
Mortgage Term
Tranche Price
Tranche Std
Plain Vanilla MBS
10 15 20 25 30
Tranche A 359,087 360,544 361,634 362,935 363,449
Tranche B 1 213,692 1 216,450 1 218,640 1 220,795 1 221,831
Tranche C 300,735 299,524 297,814 297,235 296,145
Tranche D 158,501 160,278 161,276 161,994 162,407
Tranche A 1 348,792 1 349,485 1 350,247 1 350,328 1 351,102
Tranche B
Tranche C
Tranche D
Total
I I I I I
Tranche A 1 2,530 1 3,427 1 4,121 1 4,471 1 5,064
Tranche B 1 3,111 1 3,713 1 4,447 1 4,832 1 5,543
Tranche C 1 8,399 1 9,632 1 10,566 1 10,701 1 10.933
Tranche D 1 4,975 1 5,596 1 6,035 ( 6,230 1 6,262
Total 17,554 20,665 23,520 24,388 25,978
Tranche A
Tranche B
Tranche C
Tranche D
Total
Table 7 above presents the tranche prices in different mortgage term. As expected,
tranche A has smaller standard deviation if we take principal magnitude into account. It can be
explain by the fact that senior tranches are retired at earlier period when interest rate is still away
from the long term equilibrium rate, so people are pretty sure whether they should prepay or not.
Surprisingly, all tranche prices are growing with the increase of mortgage term except tranche C.
Besides, tranche C has the largest standard deviation, almost twice as large as tranche A. These
findings may reveal the fact that tranche C's payments are in the most unstable period and many
elements impact the price of that tranche.
8.2.2 Change CIR Parameters
Table 8 Tranche Price (Different Mean Reverse Speed)
Tranche Price
Tranche Std
Plain Vanilla MBS
Mean Reverse Speed
r(0)=5%
Tranche A
Tranche B
TrancheC
Tranche D
Total
r(0)=10%
0.04
363,011
218,311
304,673
166,310
1,052,304
Tranche I3
Tranche C
Tranche D
Total
0.07
363,577
220,967
299,594
163,878
1,048,Ol 5
9,979
22,699
13,267
53,384
0.1 6
361,508
221,071
293,784
161,248
1,037,611
0.1
363,449
221,831
296,145
162,407
1,043,832
7,142
15,466
8,880
36,844
0.13
362,738
221 ,998
295,061
161,892
1,041,689
5,601
10,395
6,027
26,041
4,702
8,537
4,952
21,570
4,190
7,080
3,997
18,453
Table 9 Tranche Price (Different Interest Volatility)
Plain Vanilla MBS
nterest Volatility I 1 0.0025 1 0.0125 1 0.0225 1 0.0325
Tranche Price
Tranche A 1 364,005 1 363,813 1 363.449 1 361,758
I Tranche B 1 222,955 1 222,762 1 221,831 1 219,340 --
r(0)=5% I Tranche C 1 295,591 1 295,899 1 296.145 1 297,417
I Tranche D ( 162,255 1 162,360 1 162.407 ( 162,745
I Total 1 1,044,806 1 1,044,834 1 1.043,832 1 1,041,260
I Tranche A 1 351,434 1 351,310 1 35lIlO2 1 350,454
r(0)=10% I Tranche C 1 284,384 1 284,339 1 284.698 1 286,324
Tranche €3 21 4,749
Tranche D
Total
Tranche A
Tranche B
Tranche Std
21 4,458
156,120
1,006,688
r(0)=5%
600
654
r(O)=lO% I Tranche C 1 1,214 1 5,926 1 10,395 1 14,939
21 3,627
156,034
1,006,141
Tranche C
Tranche D
Total
Tranche A
Tranche B
212,184
2,994
3,274
Table 8 and 9 examine the CIR parameter effect on the tranche price and standard
deviation. Compared to junior tranches, senior tranches are less influenced by the change of
interest volatility and mean reverse speed. Their standard deviations change slower than the
junior ones which accord with the thought that senior tranches have comparatively stable cash
flow streams.
156,239
1,005,666
1,221
716
2,975
719
707
Tranche D
Total
156,474
1,005,435
5,064
5,543
7,069
6,840
6,109
3,492
14,846
3,505
3,404
694
3,138
10,933
6,262
25,978
5,961
5,601
3,361
15,208
14,530
8,292
33,612
8,261
7,335
6,027
26,041
8,603
36,211
CONCLUSION
Our paper provides a big picture of mortgage-backed securities and a general idea of how
to price mortgage-backed securities. In the paper, we use reduced form method to pricing MBS
and CMOS. Interest rates were simulated by using CIR model. A set of cash flow was generated
by the prepayment model and final price is computed by discounting all the cash flows. The
numerical results presented help us determine how risk is allocated in mortgage pool and each
tranches.
The results reveal that the price of mortgage pool is unconditionally positive related with
mortgage term and coupon rate and negative related with interest volatility. When we change the
interest reverse speed, the pool price gives us two different tendencies. It goes up with the
increase in the reverse speed, if we set initial rate higher than mortgage coupon rate and falls if
we set initial rate lower than mortgage coupon rate. Mortgage's standard deviation goes up with
the increase of mortgage term, coupon rate and interest volatility. It rocket even we change
volatility by 1 percent. Its risk declines if we increase the interest reverses speed, which could be
explained by the more predictable cash flows.
Things become more complex when we deal with a plain vanilla structure MBS. The
numbers still realize some tendencies of the change of the tra~che price and tranche standard
deviation. However, those tables reveal that risks have been redistributed in tranches in some
forms and could be studied by changing the coefficients. In this paper, I only assume a plain
vanilla structure; however, my simulations can be easily implemented to any other type of CMO
and more complicated CMO structures could be studied by this method.
REFERENCE LIST
Akesson, F. and J Lechoczky. "Path Generation for Quasi-Monte Carlo Simulation of Mortgage Backed Securities." Management Science, Vo1.46, No.9 (2000)
Bandic, I. "Pricing Mortgage-backed Securities and Collateralized Mortgage Obligations." M.sc. Essay (2004), University of British Columbia.
Buser, S.A. and Hendershott, P.H. "The Pricing of Default-Free Mortgages." Working paper No. 1408 (1984), National Bureau of Economic Research.
Cox, C., J. E. Ingersoll and S. A. Ross. "A theory of the term structure of interest rate." Econometrics, Vo1.53 (1985): 385-407.
Durn, K. B. and Chester S. Spatt. "The Effect of Refinancing Costs and Market Imperfections on the Optimal Call Strategy and the Pricing of Debt Contracts." Working paper (1986), Carnegie-Mellon University.
Durn, K. B. and J. McCornell. "A Comparison of Alternative Models for Pricing GNMA Mortgage-Backed Securities." Journal of Finance, Vo1.36 (1981): 471-483.
Dunn, K. B. and J. McCornell. "Valuation of Mortgage- Backed Securities." Journal of Finance, V01.36 (1981): 599-617.
Fabozzi, Frank J. and Modigliani Franco. Mortgage and Mortgage-Backed Securities Markets. Boston: Harvard Business School Press, 1997
Fabozzi, Frank J. Fixed Income Mathematics Analytical & Statistical Techniques. Chicago: Irwin Professional Publishing, 1997
Fabozzi, Frank J. The handbook of Mortgage Mortgage-Backed Securities. Chicago: Probus Publishing, 1992
Fabozzi, Frank J., Andrew J. Kalotay and George 0 . Williams. "A Model for Valuing Bonds and Embedded Options." Financial Analysts Journal (1993): 35-46.
Hull, J. Options, Futures, & Other Derivatives. Upper Saddle River: Prentice Hall, 2000
Richard, S. and R. Roll. "Prepayment on Fixed-Rate Mortgage-Backed Securities." Journal of Portfolio Management, Vo1.15 (1 989): 73-82.
Schwartz, E. and W. Torous. "Prcpayment and the Valuation of Mortgage-Backed Securities." Journal of Finance, Vo1.44 (1 989): 375-392.
Stanton, R. "Rational Prepayment and the Valuation of Mortgage-Backed Securities." The Review of Financial Studies, Vo1.8, No.3 (1995): 667-708.
Tirnrnis, G. C. "Valuation of GNMA Mortgage-Backed Securities with Transaction Costs, Heterogeneous Households and Endogenously Generated Prepayment Rates." Working Paper (1 985), Carnegie-Mellon University.
Vasicek, 0. "An Equilibrium Characterization of the Term Structure." Journal of Financial Economics, Vol. 5 (1977): 177-1 88.
Wallace, N. "Innovations in Mortgage Modelling: An Introduction." Real Estate Economics, 33(4): 587-593